Iperf suggestion


>Date: Sun, 3 Nov 2002 10:47:51 -0500
>From: "Raghurama 'REDDY'" <rreddy --at-- psc.edu>
>Reply-To: rreddy --at-- psc.edu
>To: IPERF-USERS --at-- DAST.NLANR.NET
>CC: RREDDY --at-- vms.psc.edu, COTTRELL --at-- SLAC.STANFORD.EDU
>Message-Id: <021103104751.20c0d46d --at-- psc.edu>
>Subject: Iperf suggestion
>X-MD5SUM: 6148d5ed2bc48617e3656bc763c46498
>
>Hi,
>
>if you are considering these changes, please also consider the following
>that will be useful for network testing:
>
>IPERF_PRE_WAIT_MSEC:
>
>    sugessted below already
>	
>IPERF_POST_WAIT_MSEC
>
>     marginally useful; could do it for consistency.
>
>IPERF_INTER_WAIT_MSEC 
>
>    useful for simulating a real application that sends for a while,
>    does work, etc.  this should apply for the receiver as well as for
>    the sender.
>
>the last idea came from the book Steven's TCP/IP book; it had a program
>called "sock" that had a number of features, and i thought these would
>be particularly useful.
>
>thanks!
>
>--rr
>
>From:	SMTP%"cottrell --at-- SLAC.Stanford.EDU"  "Cottrell, Les"  3-NOV-2002
09:51:15.16
>To:	iperf-users --at-- dast.nlanr.net
>CC:	rreddy --at-- psc.edu
>Subj:	Iperf suggestion
>
>To more elegantly use web100 with iperf, Raghurama 'REDDY'
[rreddy --at-- psc.edu] has modified iperf as follows. This might be interesting
to consider as a modifiecation to the distributed version.:
>
>>From Raghurama 'REDDY' [rreddy --at-- psc.edu]:
>"i have modified iperf slightly so that it waits for some number of
seconds after connect but before actually sending out the data (this is
mainly for the client code; keep in mind this change invalidates the
transfer rates you see on the server, but the 
>clients side are correct). 
>the time to wait is determined by the env variable IPERF_PRE_WAIT_MSEC (in
msec):
>
>        setenv IPERF_PRE_WAIT_MSEC 2000
>
>this is my modification to PerfSocket_TCP.  when i run iperf, it waits for
2 seconds before starting to send the data.  within this time, i have a
script that determines the CID based on the port.
>
>here are is what i usually do (getcidbyport is my script that gets the
most recent connection to port 5821, and it waits for 1 second before
attempting to get the CID giving iperf some time to get connected):
>
>------------
>kirana% setenv IPERF_PRE_WAIT_MSEC 2000
>kirana% iperf -c vh -p 5821 -w 1m & ; gutil vdt ThruBytesAcked
`getcidbyport 5821 1`            
>[1] 10032
>------------------------------------------------------------
>Client connecting to vh, TCP port 5821
>TCP window size:  2.0 MByte (WARNING: requested  1.0 MByte)
>------------------------------------------------------------
>Waiting for 2000 msec before starting write ...done
>[  3] local 192.88.115.178 port 36001 connected with 128.182.66.118 port 5821
>[ ID] Interval       Transfer     Bandwidth
>[  3]  0.0-10.0 sec  95.5 MBytes  79.7 Mbits/sec
>-------------
>
>hope this helps!"
>
>
--
Mitch Kutzko | mitch --at-- dast.nlanr.net | mitch --at-- ncsa.uiuc.edu | 217-333-1199
http://hobbes.ncsa.uiuc.edu/


--
Mitch Kutzko | mitch --at-- dast.nlanr.net | mitch --at-- ncsa.uiuc.edu | 217-333-1199
http://hobbes.ncsa.uiuc.edu/



Other Mailing lists | Author Index | Date Index | Subject Index | Thread Index